TalkChess.com

cc2150dx wrote: ↑

Tue Jan 28, 2020 2:35 pm

Off topic:

Is "Fpu" the same as "FpuValue" ? because I don't see "Fpu" in the uci options. Thanks

Yes, the same.

Do we have now new LC0 default settings? The tuned settings were done using a mix of short time controls using the T58 (58613) network. These settings should be approximately optimal for all time controls with <20k nodes / move. Even though the CPuct scaling curve is not uniquely determined on slow...

Yeah, all nets run like this. It'd be cool if there was a problem and solving it would speed up Leela (to levels perhaps even higher than your phone!?) Something noticeable is that it starts only using a single CPU, it takes a long while to start using the 4 threads. Do these settings look fine? Pe...

Okay, thanks, that's really weird. My output comes from a Q8200 2.33Ghz @ 4CPU, if my phone is faster it'll probably be because Leela hates my computer, but now I really need to check how she runs on my phone. A x250 speedup for running Leela on a device I already have?? :shock: 10 NPS with 256x20b...

You probably mean 20 seconds per move, right? Or your full CPU churns out 15 nps with the fast T59 net? My phone is much faster. Yeah, it's the slowest thing I've ever seen: CPU Leela Openblass 256x20-mod-swa-1541000.pb network: 1/02 0:02 +0.32 1.e4 e5 (3) 0 2/03 0:07 +0.29 1.e4 e5 2.Nf3 (5) 2/04 0...

To check the time control behavior, I used three time controls: 6s + 0.1s, corresponding to about 20k npm on average, 30s + 0.3s, corresponding to about 100k npm and 120s + 1.2s, corresponding to about 400k npm on average. I used a late T59 net in self-games Kiudee versus default. Here are the resu...

The smaller than naive error margins come from 5-nomial variance for unbalanced openings , discussed here: https://www.chessprogramming.org/index.php?title=Match_Statistics&mobileaction=toggle_view_desktop#Statistical_Analysis However one may show that under reasonable elo models the trinomial mode...

It's good that you opened this thread. Yes, Kiudee settings are good within a large range of generality. I will present some current results with one the latest T59 nets on my OC-ed RTX 2070 GPU, these 128x10b nets with Lc0 0.23.2 hit the ceiling of about 100k NPS of the engine's (pseudo) MCTS searc...

Thanks for the info, I can confirm at 30'' + 0.3'' in 200 games from unbalanced openings: Score of SV_256x20_1541_Kiudee vs SV_256x20_1541: 52 - 30 - 118 [0.555] 200 Elo difference: 38.37 +/- 30.78 Finished match On Twitch (potato_chess), 2 testruns with Net 62038 were played (1'+2'' on RTX, which ...

Of course the people that “know”, know that a big grid search for parameters at different time controls is expensive, long winded and not fun. That’s why these “special settings” only come along every once in a while, probably aren’t best or even good at most time controls, and are passed around li...